See what I love about Matte Wolfe is that in comparison to another TH-camrs he reads the room better. He looks at this from a perspective of an average user, and explains stuff from A to Z in a very understandable language with examples. Great work!
Thanks Matt for a great explanation. For those of you considering giving this a go, it's worth mentioning: 1) You need to enter some payment details and buy some API credits. I bought $10 worth. Minimum is $5. 2) I tried the example similar to Matt's. "Find out the weather (min max temps) for the next few days for my locality and put them into a spreadsheet." 3) DOCKER had to install Firefox as it was the first time. 4) I exceeded my maximum API calls per minute limit (for Tier 1 users). 5) When I asked Anthropic to increase my rate limit (via their form) they don't accept a Gmail account as an email address (even though I signed up, like Matt, via my Google account) 6) That failure cost 19c.
Do I have to enter my company details to buy credit? When I tried to buy credits it asked so much information about how my company is going to use it, as like I am adding a corporate payment
lol there won't be any jobs in 5 years. Remember 5 years ago was October 2019. That was still over 3 years before ChatGPT and this modern AI boom, and it was 6 months before Covid. 5 years is a lot of time and AI could replace all jobs in the next 5 years.
Woahh this looks really cool and I can see the massive potential it has! The people who are watching your channel and other AI news channels keeping up with what's coming out as it comes out will be the ones ahead of the game! This agent is still a baby but in a year or so I can see this being something that everyone is using to really get their work/ideas really going. Exciting times ahead!
This is great stuff Matt, thanks for putting this together. I asked it to do a simple web search for a local golf club, find some tee times and put them in a spreadsheet. It failed multiple times, including rate limit failures. I then gave the actual website address directly and it failed to find it again. It cost ~$.8 per request failure. It still could not access the website. Cool tech, but still a ways to go before its useable
very good points. My interest is the point when it becomes so polished that it is a package or exe that we can install on our own PC's, and "turn it loose" so to speak, to operate our own desktops.
Matt I love your channel, I tried this same prompt you used at the beginning of the video, and every single time I ran it multiple times,I got the same Token runtime error, even after increasing the limit, while it may be the future, it's not at the present yet when it comes to any type of consistency.
Hahaha you got me so excited in the first minute XD But it's not interesting enough to already use right now i think. But give it a couple of months and this will be amazing! Especially when we can start automating with it 😍
They say right in the video demo, it'll be improving a lot of over the next couple of months, and you won't have to babysit it and walk it through all these steps but rather it will be able to do everything end-to-end.
Kudos to Matt Wolfe: He does actually test things and show it. So many others just take 'demos' from the vendors and show that of but show also the fails of the tools. Its very interesting technology. I did create lots of automation tools over the last 20 years and this is very interesting. I wonder just how much it will cost to run these tools, as the credits for Matt Wolfe cost him as $1.50. So real work flows, they seem they can add up real high. My own automation tools (non-ai), would do about 300 times the 'api calls' but that was based on templating, ML inference, API's and so on, which is of course cheaper but that a lot more to develop and get running. Hope Matt Wolfe can post some more dirrect workflow videos to see how it would work in a real scenario. Thanks for the great video!
They need to partner / merge with a Robotic Process Automation software maker like Automation Anywhere or MS Power Automate, etc. They already have the tools to do the software / searching and web scraping stuff (and it would use less tokens). Then they can combine it with their higher level functions rather than reinventing the RPA wheel. RPA tools are great but can be complex and laborious to set up; a merger between them would be gold mine and a godsend.
@@SocratesWasRight There is a product called Leapwork that integrates with Chatgpt, but its focus is more on Test Automation but we use it as an RPA tool
no need to partner specifically with anyone. they'll be able to work with any software that a human can - no need to lock into any particular application or vendor. in fact, current tools like the ones you mentioned may even be disintermediated. sorry.
Matt, I have been watching your channel since the beginning (or nearly since then). I love seeing your success, passion and creativity! Always great content!
This is so cool, I would literally need something like this right now to fill in the details for 40+ digital products (title, description, keywords, etc) as currently I have to do all that frustrating stuff manually. Its def still in its early baby steps, but i do see the massive potential. Ideally this kind of functionality should be part of the OS, running locally with a low-mid range hardware. That's a future I'm really excited for; I can be the director/product owner and the implementation/grunt work would be done by the AI.
Pfff, now I'll finally have to install Docker. For some reason it didn't want to install on my Manjaro but now I have no choice... Claude is amazing. I really love what Anthropic are doing.
Great video again. Tried the agent it out but it seems to be extremely expensive. I only got to the point where the agents got the weather overview and I was down 18 cents already! This was even before the spreadsheet was created. So sending these agents to work and do complex tasks will cost you a lot, and I mean a lot !!! of money obviously.
I was all quite excited yet, ignorant about agents. This has been an excellent presentation, in that, now I'm not so excited but looking forward to how this is going to develop.
I remember the wonderful days of hand coding web sites, doing a find and replace on an entire web site. The computer would open all the pages and make the coding changes and save the pages one by one, faster than you could track what it was doing. It was very magical. Like the player piano of computers ;-)
This is incredible! 👀 Remote Tech support jobs will be taken over with tech like this! Plus so many other things. The potential uses so vast. I will definitely be doing this. My virtual ai employees just got an upgrade. 😁
I’m looking forward to the day when the agents can look at your video or anybody’s TH-cam video and just do all the steps that you’re showing without us having to actually do all that stuff. I would imagine that’s not many generations away from what we’ve got now. with all the transcription and screenshots and pretty sure I could figure out exactly what a person is showing on a video and duplicate those steps, saving us tons of work.
Now I understand why I can't find the AI tool I need they don't have it yet but this looks like a good start. Thank you Matt! I am looking for a tool that can do everything in my computer I tell it to. I think I can already do that in Windows I am looking into it now.
I agree, this is going to be very useful, perhaps soon. My 1st thought is "Search all my storage locations and find the .jpg's and copy only 1 of each pic to a backup drive xxx", which puts everything in one place, eliminates dups, and backs up my pics. It may hang due to too many pics or drives so I would then cut the request into smaller bites. Still beats the tedium of me doing it as I have been procrastinating doing. This I would pay for (when done) and I am a devoted cheapscake when there is so much free on the internet.
I just want my own AI agents to help me get more work. I want it to go out there and find the right clients for me and help me make money to get by. As soon as I can figure out how to do that I will try.
Could describe your target clients. If your clients are people in business in your area, you could get it to make spreadsheet for potential clients and contact details. You could even have it execute cold call emails :)
@@juicegod777Not true. For example, as a developer, I had tools for LLMs to control my computer long before they came out. If you have the drive to do it yourself before it’s easier, you can get opportunities before others do
It does have some parallels to *Ghost in the Shell*, especially with the concept of AI agents performing tasks autonomously. In the *Ghost in the Shell* universe, there's a deep exploration of the integration of technology and humanity, where cybernetic enhancements and AI play significant roles in everyday life. The video you watched highlights a similar idea-using AI to handle tasks and navigate the digital world, almost like a digital assistant or extension of oneself. The advancements in AI showcased, like navigating browsers and creating spreadsheets, reflect a growing trend towards integrating AI into personal and professional tasks, reminiscent of the themes in *Ghost in the Shell*. As AI continues to evolve, it raises interesting questions about autonomy, identity, and the nature of intelligence, just like in the anime. How do you see this development influencing our future interactions with technology?
Very cool, but seems like the demo is just a prompt based scraper. What I am personally curious about is if it can interact with a CRM. If it can provide the same control and prompt based interface to manage/run/use a CRM, it’s a different world.
Good job Matte. I love that you can use Docker and still appear as the "average User". So now I wonder if you actually understand the docker call parameters :)
Thanks for this Matt. I have tried it but at the terminal i got this instead of the link: docker: unexpected eof. Has anyone run into this and how did you solve it? Please help a non-techie
Hey @Matt, thank you so much for your work!! I've been following you since the beginning, and honestly, you`re the only one I follow continuously every week. Could you tell me how to solve the cookie problem with the Firefox browser that the agent is opening? My agent is stuck there.
Totally out of curiosity, is there a way to do it out of the VM, meaning directly in the desktop? I have a blank PC and just wondered. Haven't seen it yet.
Using Imac M1. I tried everything up to 3:33 but on pressing Enter, nothing happens. I have put in my own API key generated by my Anthropic but nothing happens. Any clues?
Thanks, but I receive an error message after copying the API code in: "docker: invalid reference format: repository name (library/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) must be lowercase." what does this mean?
With my kids finishing university and high school next year, I wonder what future they'll have. I think predictions of mass job losses next year are wrong, but in 5 or 10 years? If this is the first (near) agent available, then by the time they finish there will be far better ones being used everywhere
Hi Matt, thanks for sharing this interesting software. I have some issues when launching the localhost, when i am trying to use the same propts as you did it shows that Error code: 400 - {'type': 'error', 'error': {'type': 'invalid_request_error', 'message': 'Your credit balance is too low to access the Anthropic API. Please go to Plans & Billing to upgrade or purchase credits.'}}. I thought there was a free trial version but still doesnt allow me to test the software. Any clue?
Amazing. But like with what all the other AI guys are saying the tokens are way expensive to use and it is slow. Perhaps in a few months it will be affordable once others publish their own.
they could make the thing much more efficient if they took fewer screenshots and batched keys... for example, entering all the Excel data and arrow presses before uploading another screenshot.
I can see a future version where there's a common interface with applications and the app builder, much like the menu systems they have, also has an AI tasks that's internal to the app and this common interface references that to control the application. That way, the AI doesn't need to know how to use every version of every application ever built. Maybe we'll be seeing that from Microsoft soon.
so since its on a virtual machine you cant get the files it saves? if you had it make a spreadsheet and save it? Is it able to login to sites for you like say you asked it to go into gmail and unsubscribe from all your spam emails.
I keep getting this response from Docker: invalid reference format: repository name (/Users/mac/.anthropic) must be lowercase. See 'docker run --help'. what do i do?
Thank you for the Agent introduction. Wouldn’t it be more helpful if the Agent could create a Python RPA script that does what you want? Hence token/process limits wouldn’t be an issue
See what I love about Matte Wolfe is that in comparison to another TH-camrs he reads the room better. He looks at this from a perspective of an average user, and explains stuff from A to Z in a very understandable language with examples. Great work!
He shows better use cases too. And he draws from good sources and doesn't go stinky rants
Facts.
I see
Makes sense. I'm like: Who doesn't have docker installed already? Why is he using Windows?
@@nathanbanks2354why would he NOT use Windows when it is the most used OS in the world?
Thanks Matt for a great explanation. For those of you considering giving this a go, it's worth mentioning:
1) You need to enter some payment details and buy some API credits. I bought $10 worth. Minimum is $5.
2) I tried the example similar to Matt's. "Find out the weather (min max temps) for the next few days for my locality and put them into a spreadsheet."
3) DOCKER had to install Firefox as it was the first time.
4) I exceeded my maximum API calls per minute limit (for Tier 1 users).
5) When I asked Anthropic to increase my rate limit (via their form) they don't accept a Gmail account as an email address (even though I signed up, like Matt, via my Google account)
6) That failure cost 19c.
Do I have to enter my company details to buy credit? When I tried to buy credits it asked so much information about how my company is going to use it, as like I am adding a corporate payment
When it works, it works though! Any luck lately or you tried just the once?
Jobs will look wildly different 5 years from now.
Jobs?
lol there won't be any jobs in 5 years. Remember 5 years ago was October 2019. That was still over 3 years before ChatGPT and this modern AI boom, and it was 6 months before Covid. 5 years is a lot of time and AI could replace all jobs in the next 5 years.
the world
@@vectoralphaSec SERIOUS!
Yes - he'll be much more rotted...
Woahh this looks really cool and I can see the massive potential it has! The people who are watching your channel and other AI news channels keeping up with what's coming out as it comes out will be the ones ahead of the game! This agent is still a baby but in a year or so I can see this being something that everyone is using to really get their work/ideas really going. Exciting times ahead!
This is great stuff Matt, thanks for putting this together. I asked it to do a simple web search for a local golf club, find some tee times and put them in a spreadsheet. It failed multiple times, including rate limit failures. I then gave the actual website address directly and it failed to find it again. It cost ~$.8 per request failure. It still could not access the website. Cool tech, but still a ways to go before its useable
very good points. My interest is the point when it becomes so polished that it is a package or exe that we can install on our own PC's, and "turn it loose" so to speak, to operate our own desktops.
Matt I love your channel, I tried this same prompt you used at the beginning of the video, and every single time I ran it multiple times,I got the same Token runtime error, even after increasing the limit, while it may be the future, it's not at the present yet when it comes to any type of consistency.
Hahaha you got me so excited in the first minute XD But it's not interesting enough to already use right now i think. But give it a couple of months and this will be amazing! Especially when we can start automating with it 😍
They say right in the video demo, it'll be improving a lot of over the next couple of months, and you won't have to babysit it and walk it through all these steps but rather it will be able to do everything end-to-end.
GG'S FELLOW HUMANS WE HAD A GOOD RUN DIDN'T WE? IT WAS FUN WHILE IT LASTED
I can draw a stick figure hehehehe....
@@nathanbanks2354yeah stupid ai can’t even draw a stick figure we’re clearly superior lol
Boohoooo
@@nathanbanks2354a while ago ChatGPT couldn’t figure out how many “r” in “strawberry”
If your job can be replaced by AI it's not a real job
Is there an option to turn the leaking faucet sound off?
😆
Volume down lol
Thank you for the tutorial. This was written with Technosapien's new Claude Agent - P.S. I am a big fan
Absolutely INSANE! Things are moving at such a fast pace, amazing times...
Kudos to Matt Wolfe: He does actually test things and show it. So many others just take 'demos' from the vendors and show that of but show also the fails of the tools. Its very interesting technology. I did create lots of automation tools over the last 20 years and this is very interesting. I wonder just how much it will cost to run these tools, as the credits for Matt Wolfe cost him as $1.50. So real work flows, they seem they can add up real high. My own automation tools (non-ai), would do about 300 times the 'api calls' but that was based on templating, ML inference, API's and so on, which is of course cheaper but that a lot more to develop and get running. Hope Matt Wolfe can post some more dirrect workflow videos to see how it would work in a real scenario. Thanks for the great video!
They need to partner / merge with a Robotic Process Automation software maker like Automation Anywhere or MS Power Automate, etc. They already have the tools to do the software / searching and web scraping stuff (and it would use less tokens). Then they can combine it with their higher level functions rather than reinventing the RPA wheel.
RPA tools are great but can be complex and laborious to set up; a merger between them would be gold mine and a godsend.
Yeah... LLMs being frontends for RPA/other similar tools.
@@SocratesWasRight There is a product called Leapwork that integrates with Chatgpt, but its focus is more on Test Automation but we use it as an RPA tool
no need to partner specifically with anyone. they'll be able to work with any software that a human can - no need to lock into any particular application or vendor. in fact, current tools like the ones you mentioned may even be disintermediated. sorry.
@@SocratesWasRightYES! Great call!
Thanks!
Greetings Mr. Matt. You are just great. I appreciate all you do for us. 👏👏👏
thanks for always taking the time to do this!
Matt, I have been watching your channel since the beginning (or nearly since then). I love seeing your success, passion and creativity! Always great content!
This is so cool, I would literally need something like this right now to fill in the details for 40+ digital products (title, description, keywords, etc) as currently I have to do all that frustrating stuff manually.
Its def still in its early baby steps, but i do see the massive potential.
Ideally this kind of functionality should be part of the OS, running locally with a low-mid range hardware.
That's a future I'm really excited for; I can be the director/product owner and the implementation/grunt work would be done by the AI.
I am so glad that I found your channel. By the way, your AI voice in the videos is really good. Keep up the good work!
Thanks Matt, for the detailed instructions. I hope you will update us on this application as it develops!
Pfff, now I'll finally have to install Docker. For some reason it didn't want to install on my Manjaro but now I have no choice... Claude is amazing. I really love what Anthropic are doing.
This is the first time openAI isn't the company that did something new with AI. Curious on how they will compete.
Great video again. Tried the agent it out but it seems to be extremely expensive. I only got to the point where the agents got the weather overview and I was down 18 cents already! This was even before the spreadsheet was created. So sending these agents to work and do complex tasks will cost you a lot, and I mean a lot !!! of money obviously.
Yeah-same here! Mine failed (max API calls exceeded) too.
@@Just4Growers exactly my experience!
That's crazy haha "Go Max Out My Character On Old School Runescape"
hahahaha letsssss go inflation in games
currently it appears your using a desktop that they create and provide, not your own
I was all quite excited yet, ignorant about agents. This has been an excellent presentation, in that, now I'm not so excited but looking forward to how this is going to develop.
Wow! This is AMAZING!!!
Thanks, Matt! Very exciting!
Matt number 1. You are the best. And this news is big!
I remember the wonderful days of hand coding web sites, doing a find and replace on an entire web site. The computer would open all the pages and make the coding changes and save the pages one by one, faster than you could track what it was doing. It was very magical. Like the player piano of computers ;-)
Valeu!
Finally, AI can delete my browser history when I'm gone.🥺🥺
Thank you AI.
Now we need a team of agents that talk to each other and carry out tasks on the computer 😀
This is incredible! 👀
Remote Tech support jobs will be taken over with tech like this! Plus so many other things.
The potential uses so vast. I will definitely be doing this. My virtual ai employees just got an upgrade. 😁
I’m looking forward to the day when the agents can look at your video or anybody’s TH-cam video and just do all the steps that you’re showing without us having to actually do all that stuff. I would imagine that’s not many generations away from what we’ve got now. with all the transcription and screenshots and pretty sure I could figure out exactly what a person is showing on a video and duplicate those steps, saving us tons of work.
Now I understand why I can't find the AI tool I need they don't have it yet but this looks like a good start. Thank you Matt! I am looking for a tool that can do everything in my computer I tell it to. I think I can already do that in Windows I am looking into it now.
I agree, this is going to be very useful, perhaps soon. My 1st thought is "Search all my storage locations and find the .jpg's and copy only 1 of each pic to a backup drive xxx", which puts everything in one place, eliminates dups, and backs up my pics.
It may hang due to too many pics or drives so I would then cut the request into smaller bites. Still beats the tedium of me doing it as I have been procrastinating doing.
This I would pay for (when done) and I am a devoted cheapscake when there is so much free on the internet.
This video gave me such a clear understanding!
FIRE 🔥
Game changer. Bouta go pick up 10 remote jobs
Bingo
I just want my own AI agents to help me get more work. I want it to go out there and find the right clients for me and help me make money to get by. As soon as I can figure out how to do that I will try.
Could describe your target clients. If your clients are people in business in your area, you could get it to make spreadsheet for potential clients and contact details. You could even have it execute cold call emails :)
@@thomasrea8648 Heh, we're at the point it can execute cold calls themselves. Realtime API, Hume API, etc. + Twilio.
Bro by the time you figure that out, every company will be using the same tech
@@juicegod777Not true. For example, as a developer, I had tools for LLMs to control my computer long before they came out. If you have the drive to do it yourself before it’s easier, you can get opportunities before others do
just one question. all these steps were executed in your local machine (local browser) or inside a docker container ?
It does have some parallels to *Ghost in the Shell*, especially with the concept of AI agents performing tasks autonomously. In the *Ghost in the Shell* universe, there's a deep exploration of the integration of technology and humanity, where cybernetic enhancements and AI play significant roles in everyday life. The video you watched highlights a similar idea-using AI to handle tasks and navigate the digital world, almost like a digital assistant or extension of oneself.
The advancements in AI showcased, like navigating browsers and creating spreadsheets, reflect a growing trend towards integrating AI into personal and professional tasks, reminiscent of the themes in *Ghost in the Shell*. As AI continues to evolve, it raises interesting questions about autonomy, identity, and the nature of intelligence, just like in the anime. How do you see this development influencing our future interactions with technology?
Very cool, but seems like the demo is just a prompt based scraper.
What I am personally curious about is if it can interact with a CRM. If it can provide the same control and prompt based interface to manage/run/use a CRM, it’s a different world.
This video should have 20 million views.
Good job Matte. I love that you can use Docker and still appear as the "average User". So now I wonder if you actually understand the docker call parameters :)
We'll need to get really good at writing prompts for this to be effective. It's VERY impressive though.
How much did it cost you to do 1 task
It seems like $.67 from the quick view of the Anthropic Dashboard
@@danielpgreen 67 DOLLARS?
@@danielpgreen the rich have an advantage as always
@@jksdo88 No, that's cents $00.67 but yeah this probably gets expensive just from all of the images it's taking/reading
@@danielpgreen You can reduce the images, but yes, it gets pricey over time. Two tasks for me were around $3. Depends on the task.
Except I barely know what my job is and neither does anyone else, including my boss, so that’s a problem.
😂😂😂😂🎉
@@fromduskuntodawn .
You could always set it to "do stuff"
Keep at it while you can, you won't be replaced by AI until someone finds out
🤯 I'll be training it to fire me.
Thanks as always Matt! 💯
Thanks for this Matt. I have tried it but at the terminal i got this instead of the link: docker: unexpected eof. Has anyone run into this and how did you solve it? Please help a non-techie
Hey @Matt, thank you so much for your work!! I've been following you since the beginning, and honestly, you`re the only one I follow continuously every week.
Could you tell me how to solve the cookie problem with the Firefox browser that the agent is opening? My agent is stuck there.
can this be used within your own applications? if I trained it ?
Totally out of curiosity, is there a way to do it out of the VM, meaning directly in the desktop? I have a blank PC and just wondered. Haven't seen it yet.
Matt you've gotta put something on in San Diego. I'd love to learn TH-cam strategies from one of the best!
Using Imac M1. I tried everything up to 3:33 but on pressing Enter, nothing happens. I have put in my own API key generated by my Anthropic but nothing happens. Any clues?
Does it only allow you to use Firefox or can you use chrome? Also, does it only work with certain existing apps or can we use apps like notion?
Didnt know it created its own workspace computer, nice! I thought it was running on your local pc haha
Can it only be used like this in Docker with this virtual environment or can it take over a Windows 11 desktop right now?
This is really the beginning. We will a personal jarvis in couple years
I'm totally lost. Where did you get that line of code in Docker? I'm thinking you skipped a step.
How would I use this feature in a non demo setting? Or how do i adapt what apps are being used?
Hallelujah FinALLY! There's other methods but we needed something user friendly
Thanks, but I receive an error message after copying the API code in: "docker: invalid reference format: repository name (library/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) must be lowercase." what does this mean?
Does OpenAI have an equivalent version of Claude's Sonnet AI Agent's?
Thanks the explanation this topic. Can you also make this same topic for AgentExe & Open Interpreter? Their also can use for computer use.
Which programs Can i use only paint And Fire fox or can i Use it In render Architecture
Crikey! Amazing agent.
Is it using your locaol desktop or a cloud computer desktop??
is the raindrop sound the default?
AI is the silent hero of modern problem-solving ✨
With my kids finishing university and high school next year, I wonder what future they'll have. I think predictions of mass job losses next year are wrong, but in 5 or 10 years? If this is the first (near) agent available, then by the time they finish there will be far better ones being used everywhere
Thank you so much for sharing. This is amazing ！
Solid practical content man 👍
Thabk you Wolfe,
Great work!
Can it interact with applications running in my windows... Like what about a minor edit in my premiere pro? Or it only works in browser ?
It might if you vnc to your windows machine
Another banger!
Question: how safe is it to use your API Key on a virtual computer?
after u close docker how do u run this again? do we have to paste in the terminal to restart?
Hi Matt, thanks for sharing this interesting software. I have some issues when launching the localhost, when i am trying to use the same propts as you did it shows that Error code: 400 - {'type': 'error', 'error': {'type': 'invalid_request_error', 'message': 'Your credit balance is too low to access the Anthropic API. Please go to Plans & Billing to upgrade or purchase credits.'}}. I thought there was a free trial version but still doesnt allow me to test the software. Any clue?
Great knowledge 🎉
Thank you, thank you! I see that it's still a bit limited but in time... watch out!
Amazing. But like with what all the other AI guys are saying the tokens are way expensive to use and it is slow. Perhaps in a few months it will be affordable once others publish their own.
Damn....excellent weather. Its 89 today in South Florida.
First task. Copilot AI did this PERFECTLY and QUICKLY. No need for other bullcrap API setup. The image part is tough even to copilot, GPT..
Can you ask it to throttle your per minute data use to avoid the limit
Did you test this in a virtual machine or on you personal desktop?
He said he tested on a virtual machine.
@@MartinaRoters Thank you
they could make the thing much more efficient if they took fewer screenshots and batched keys... for example, entering all the Excel data and arrow presses before uploading another screenshot.
what was the total cost of your test using the API? It's a lot of visual learning tokens
I can see a future version where there's a common interface with applications and the app builder, much like the menu systems they have, also has an AI tasks that's internal to the app and this common interface references that to control the application. That way, the AI doesn't need to know how to use every version of every application ever built. Maybe we'll be seeing that from Microsoft soon.
so since its on a virtual machine you cant get the files it saves? if you had it make a spreadsheet and save it? Is it able to login to sites for you like say you asked it to go into gmail and unsubscribe from all your spam emails.
You lost me at 3:06. The API_KEY=$ snippet I can't find
How and where can I add a pdf for it to process? and Create PowerPoints based on it
Pretty soon we’re having that fully running locally
So coool! And nice explanation~
I keep getting this response from Docker: invalid reference format: repository name (/Users/mac/.anthropic) must be lowercase.
See 'docker run --help'. what do i do?
Can this be used with Google sheets?
is there any other tool besides docker? it wont load on my mac airbook m3 sonoma
How long until this works in Unity / Ue5?
Thank you for the Agent introduction. Wouldn’t it be more helpful if the Agent could create a Python RPA script that does what you want? Hence token/process limits wouldn’t be an issue
Did you get a new camera?